An affinity-modulated T cell engager targeting Claudin 18.2 shows potent anti-tumor activity with limited cytokine release

Abstract

Background: AZD5863 is a bispecific T cell engager (TCE) with high affinity to CLDN18.2 and low affinity to cluster of differentiation 3 (CD3), designed to decrease its peripheral cytokine release potential, improve the therapeutic index, and maintain potent anti-tumor activity.

Methods: AZD5863 was evaluated using CLDN18.2-expressing human cell lines alone or in co-cultures with human or cynomolgus monkey peripheral blood mononuclear cells to determine affinities, specificity, potency, and bystander killing activity. In vivo, AZD5863-mediated tumor growth inhibition and pharmacodynamics were evaluated in humanized mice or human CD3 transgenic mice implanted with CLDN18.2-expressing cancer cell lines.

Results: AZD5863 was shown to bind specifically to human and cynomolgus monkey CLDN18.2 and to CD3, with CLDN18.2 binding also conserved against the murine protein. AZD5863 mediated T cell-dependent anti-tumor activity against CLDN18.2-expressing lines, with potency significantly correlating with CLDN18.2 receptor density. Cytokine secretion induced by AZD5863, in vitro and in vivo, was lower compared with a CLDN18.2 TCE with higher affinity for CD3. AZD5863 mediated T cell-dependent bystander killing of CLDN18.2-negative cells in the presence of CLDN18.2-expressing cells, in a mechanism partly dependent on interferon gamma (IFNγ), tumor necrosis factor alpha (TNFα), and Fas ligand. In vivo, AZD5863 treatment resulted in potent tumor control in pancreatic, gastric, and esophageal models and enhanced engraftment of immune populations in a humanized model.

Conclusions: AZD5863 mediates potent anti-tumor activity in vitro and in vivo, while inducing limited levels of cytokines. This work improves our understanding of the mechanism of action of affinity balanced TCEs and informs the design of a phase 1 trial testing AZD5863 in gastric, pancreatic, and esophageal adenocarcinoma (NCT06005493).

Keywords: Bispecific T cell engager - BiTE; Cytokine release syndrome; Gastric Cancer.

Figure 1. AZD5863 binds with low affinity to CD3 and high affinity to Claudin 18.2. (A) Schematic representation of AZD5863, a bispecific antibody with two VH-only antibody domains to Claudin 18.2 (orange) and one antibody domain to CD3 (δε heterodimer, blue-green) with Fc-disabling mutations (L234A, L235A, G237A). (B) AZD5863 binds to human CD3 (δε heterodimer) protein with an affinity of 170 nM in a surface plasmon resonance (SPR) assay (representative results from three independent experiments are shown). (C) AZD5863 binds to human Claudin 18.2 VLPs (virus-like particles) with an affinity of 0.1 nM in a SPR assay (representative results of two independent experiments are shown). The fully corrected binding data for both interactions were globally fitted to a 1:1 binding model to determine the K_D using the Biacore Insight Evaluation Software (Cytiva). Light blue sensorgram trace represents the raw data and dark blue line the global 1:1 fit to that data. (D) Geometric mean fluorescence intensity (GMFI) of AZD5863 binding to cell surface–expressed human Claudin 18.2, Claudin 18.1, or to non-transduced HEK293 and Jurkat cells, determined by flow cytometry. Data from triplicates presented as mean ± SD. For EC₅₀ determinations, antibody concentration, and GMFI, data were log transformed before analysis and fit using the log agonist versus response using Prism Software (GraphPad; representative results of three independent experiments). See also online supplemental figure S1. CD3, cluster of differentiation 3.

Figure 2. AZD5863 mediates T cell-dependent cytolysis of Claudin 18.2-expressing tumor cells. (A). T cell-dependent cytotoxicity assay (TDCC) using human PBMCs from seven donors co-cultured with NUGC4 gastric cancer cells for 48 hours in the presence of a titration of AZD5863. Cytotoxicity data were normalized taking the highest and lowest values per data set as 100% and 0%, respectively. Data from triplicates presented as mean ± SD. For EC₅₀ determinations, antibody concentration and GMFI data were log transformed before analysis and fit using the log agonist versus response using Prism Software (GraphPad; representative results of three independent experiments). (B.) Comparison of EC₅₀ values of cytotoxicity, TNFa, IL2, IFNγ, and IL6 production from TDCC assay shown in (A). Statistical testing by one-way analysis of variance, with the Geisser-Greenhouse correction and Dunnett’s multiple comparisons test. Asterisks represent statistical significance (**P < 0.01; ***P < 0.001; ****P < 0.0001). (C.) Time-resolved TDCC assay comparing AZD5863 cytotoxicity and cytokine release potency across a time course assay. TDCC assay using PBMCs and NUGC4 cancer cells co-cultured for 6, 24, 48, 72, or 144 hours in the presence of a titration of AZD5863 or a control antibody (not shown). Cytotoxicity or cytokine release data was quantified over time by calculating area under the curve for individual AZD5863 concentrations and normalized by taking the highest and lowest values per data set as 100% and 0%, respectively. Data from triplicates presented as mean ± SD. For EC₅₀ determinations, antibody concentration and GMFI data were log transformed before analysis and fit using the log agonist versus response using Prism Software (GraphPad). See also online supplemental figure S2. (D.) Correlation of AZD5863 EC₅₀ of TDCC activity in multiple cancer cell lines and Claudin 18.2 expression levels (ABC/cell). Correlation testing using Pearson correlation analysis of log-transformed EC₅₀ and ABC/cell values using Prism software (Graphpad). Pearson r value and P value are shown in the graph. See also online supplemental figure S3. GMFI, geometric mean fluorescence intensity; IFNγ, interferon gamma; IL2, interleukin 2; PBMC, peripheral blood mononuclear cell.

Figure 3. AZD5863 mediates lower cytokine secretion than a CLDN18.2xCD3 T cell engager (TCE) with high CD3 affinity, in vitro and in vivo. (A) T cell-dependent cytotoxicity assay (TDCC) using PBMCs co-cultured with NUGC4 cells for 24 hours in the presence of a titration of AZD5863 or a CLDN18.2 TCE with higher affinity to CD3 (AMG910a, ie, AMG910 analog). Data from duplicates presented as mean ± SD. For EC50 determinations, antibody concentration, and GMFI data were log transformed before analysis and fit using the log agonist versus response using Prism Software (GraphPad; representative results of three independent experiments). (B) Production of IFNγ, TNFα, and IL6 from TDCC assay in response to AZD5863 or AMG910a. The experiment was repeated with n=3 independent PBMC donors, showing data from one representative donor with duplicates presented as mean ± SD. (C) Cytotoxicity and IL-6 cytokine secretion (from (A) and (B)) driven by AMG910a and AZD5863; arrows indicate delta between Emax of either measurement. (D) Concentrations of IFNγ, IL2, TNFα, and IL6 in sera from NCG mice (n=6 per group) injected with 2×10e⁷ human PBMCs, 4 hours after treatment with either AZD5863, AMG910a, or a human IgG1 isotype control. Data presented as mean±SD. Statistical testing by one-way analysis of variance with multiple comparisons to AMG910a and Dunnett’s correction, asterisks represent significance (*p<0.05; **p<0.01; ***p<0.001; ****p<0.0001). (E) Time-resolved concentration of IFNγ and TNFα in sera from NCG mice (n=6 per group, each dot indicates a single animal, mean is indicated) injected with 2×10e⁷ human PBMC 24 hours prior to treatment with either AZD5863, AMG910a, or a human IgG1 isotype control. CD3, cluster of differentiation 3; GMFI, geometric mean fluorescence intensity; IFNγ, interferon gamma; IL2, interleukin 2; PBMC, peripheral blood mononuclear cell.

Figure 4. AZD5863 mediates bystander killing of CLDN18.2-negative cells. (A) Schematic of bystander killing TDCC assay, in which human PBMCs are co-cultured with PaTu8988s sorted for high CLDN18.2 expression (WT) and prestained with CFSE cell tracker dye, mixed with PaTu8988s CLDN18.2-knock out (KO) cells, prestained with CTV cell tracker dye. WT:KO ratios are indicated. Flow cytometry is used to assess percentages of AZD5863-driven T cell dependent cytotoxicity of KO or WT cells. (B) Flow cytometry quantification of percentage of dead PaTu8988s CLDN18.2 KO cells at the indicated WT:KO ratios in the bystander killing TDCC assay, in the presence of AZD5863 titration, at 72 hours. The experiment was repeated with n=3 independent PBMC donors, showing data from one representative donor with duplicates presented as mean ± SD. (C) Flow cytometry quantification of frequencies of CD25+ and CD69+ cells out of CD4+ or CD8+ T cells, at the indicated WT:KO ratios in the bystander killing TDCC assay, in the presence of AZD5863 titration, at 72 hours. The experiment was repeated with n=3 independent PBMC donors, showing data from one representative donor with duplicates presented as mean ± SD. (D) Production of IFNγ, TNFα, and IL6 at the indicated WT:KO ratios in the bystander killing TDCC assay, in the presence of AZD5863 titration, at 72 hours. The experiment was repeated with n=3 independent PBMC donors, showing data from one representative donor with duplicates presented as mean ± SD. (E) Flow cytometry quantification of percentage of dead MDA-MB-231 (CLDN18.2-negative) cells at 30%:70% CLDN18.2+:CLDN18.2- ratio, in the presence or absence of 1 nM AZD5863 and with or without blocking antibodies for IFNγ, TNFα, or FasL, at 72 hours. Showing data from n=6 PBMC donors as mean±SD. Statistical testing by one-way analysis of variance with multiple comparisons to AZD5863; asterisks represent significance (*p<0.05; **p<0.01; ***p<0.001). CD3, cluster of differentiation 3; FasL, Fas Ligand; GMFI, geometric mean fluorescence intensity; IFNγ, interferon gamma; IL6, interleukin 6; PBMC, peripheral blood mononuclear cell; TDCC, T cell-dependent cellular cytotoxicity.

Figure 5. AZD5863 mediates control in humanized and CD3-transgenic mouse models for CLDN18.2+ tumors. (A) Tumor volumes across time of NSG mice implanted with NUGC4 tumor cells mixed in Matrigel with 1×10⁶ unstimulated PBMCs and treated with AZD5863 or isotype control, administered IP at the indicated dose (vertical dashed black lines). (B) Tumor volumes across time of NSG mice implanted with OE19 tumor cells and humanized with anti-CD3/anti-CD28 stimulated PBMCs comprising 1×10⁶ T cells (intravenous, vertical dashed purple line) and treated with AZD5863 or isotype control, administered IP at the indicated doses (vertical dashed black lines). (C). Left: tumor volumes across time of NSG mice implanted with PaTu8988sHS tumor cells and humanized with anti-CD3/anti-CD28 stimulated PBMCs comprising 1×10⁶ T cells (intravenous, vertical dashed purple line) and treated with AZD5863 or isotype control, administered IP at the indicated doses (vertical dashed black lines). Right: tumor volumes across time of NSG mice implanted with PaTu8988sHS tumor cells and humanized with anti-CD3/anti-CD28 stimulated PBMCs comprising the indicated numbers of T cells (intravenous injection, vertical dashed purple line). (D) Frequencies of different immune cell populations in response to AZD5863 or Isotype control treatments within PaTu8988sHS tumors at day 24 after humanization with anti-CD3/anti-CD28 stimulated PBMCs comprising 1×10⁶ T cell in NSG mice. P values obtained with one-way analysis of variance (ANOVA) with Dunnett’s correction for multiple comparisons to Isotype control. (E) Left: tumor volumes across time of human CD3-transgenic C57BL/6 mice implanted with MC38 tumor cells transduced to express murine CLDN18.2, treated with AZD5863 or isotype control, administered IP at the indicated doses (vertical dashed black lines). Right: tumor growth rates calculated as detailed in methods, with p values obtained with Mann-Whitney t-test, comparing each treatment group to isotype control. All tumor volume graphs showing n=6–10 mice per treatment group and mean +/- scanning electron microscopy of each measurement, with p values obtained for the last measurement with two-way ANOVA with multiple comparisons. CD3, cluster of differentiation 3; IP, intraperitoneally; PBMC, peripheral blood mononuclear cell.